Head-to-head comparison of commercial artificial intelligence solutions for detection of large vessel occlusion at a comprehensive stroke center.

Purpose: Despite the availability of commercial artificial intelligence (AI) tools for large vessel occlusion (LVO) detection, there is paucity of data comparing traditional machine learning and deep learning solutions in a real-world setting. The purpose of this study is to compare and validate the performance of two AI-based tools (RAPID LVO and CINA LVO) for LVO detection. Materials and methods: This was a retrospective, single center study performed at a comprehensive stroke center from December 2020 to June 2021. CT angiography (n = 263) for suspected stroke were evaluated for LVO. RAPID LVO is a traditional machine learning model which primarily relies on vessel density threshold assessment, while CINA LVO is an end-to-end deep learning tool implemented with multiple neural networks for detection and localization tasks. Reasons for errors were also recorded. Results: There were 29 positive and 224 negative LVO cases by ground truth assessment. RAPID LVO demonstrated an accuracy of 0.86, sensitivity of 0.90, specificity of 0.86, positive predictive value of 0.45, and negative predictive value of 0.98, while CINA demonstrated an accuracy of 0.96, sensitivity of 0.76, specificity of 0.98, positive predictive value of 0.85, and negative predictive value of 0.97. Conclusion: Both tools successfully detected most anterior circulation occlusions. RAPID LVO had higher sensitivity while CINA LVO had higher accuracy and specificity. Interestingly, both tools were able to detect some, but not all M2 MCA occlusions. This is the first study to compare traditional and deep learning LVO tools in the clinical setting.

Incidence, molecular characteristics, and imaging features of "clinically-defined pseudoprogression" in newly diagnosed glioblastoma treated with chemoradiation.

PURPOSE: Pseudoprogression (PsP) remains an elusive and clinically important, yet ill-defined, phenomena that, generally, involves a period of early radiographic progression (enhancement) followed by a period of radiographic stability or regression. In the current study, we utilized data from the control arm of a phase III clinical trial in newly-diagnosed glioblastoma to explore imaging characteristics of "clinically-defined PsP", or early radiographic progression (PFS < 6 months from chemoradiation) followed by a long post-progression residual overall survival (ROS > 12 months). METHODS: One hundred sixty-nine patients with newly-diagnosed GBM from the control arm of the AVAglio trial (NCT00943826) who presented with early radiographic progressive disease (PD) (< 6 months) were included. Clinical characteristics, topographical patterns, and radiomic features were compared between newly-diagnosed GBM exhibiting early PD and early death (< 12-month ROS, "true PD") with those exhibiting early PD and a long residual survival (> 12-month ROS, "clinically-defined PsP"). RESULTS: "Clinically-defined PsP" occurred to 38.5% of patients with early PD, and was more associated with MGMT methylation (P = 0.02), younger age (P = 0.003), better neurological performance (P = 0.01), and lower contrast-enhancing tumor volume (P = 0.002) at baseline. GBM showing "true PD" occurred more frequently in the right internal capsule, thalamus, lentiform nucleus, and temporal lobe than those with "clinical PsP". Radiomic analysis predicted "clinical PsP" with > 70% accuracy on the validation dataset. CONCLUSION: Patients with early PD that eventually exhibit "clinically-defined PsP" have distinct clinical, molecular, and MRI characteristics. This information may be useful for treating clinicians to better understand the potential risks and outcome in patients exhibiting early radiographic changes following chemoradiation.

Diffusion MRI is an early biomarker of overall survival benefit in IDH wild-type recurrent glioblastoma treated with immune checkpoint inhibitors.

BACKGROUND: Diffusion MRI estimates of the apparent diffusion coefficient (ADC) have been shown to be useful in predicting treatment response in patients with glioblastoma (GBM), with ADC elevations indicating tumor cell death. We aimed to investigate whether the ADC values measured before and after treatment with immune checkpoint inhibitors (ICIs) and the changes in these ADC values could predict overall survival (OS) in patients with recurrent IDH wild-type GBM. METHODS: Forty-four patients who met the following inclusion criteria were included in this retrospective study: (i) diagnosed with recurrent IDH wild-type GBM and treated with either pembrolizumab or nivolumab and (ii) availability of diffusion data on pre- and post-ICI MRI. Tumor volume and the median relative ADC (rADC) with respect to the normal-appearing white matter within the enhancing tumor were calculated. RESULTS: Median OS among all patients was 8.1 months (range, 1.0-22.5 months). Log-rank test revealed that higher post-treatment rADC was associated with a significantly longer OS (median, 10.3 months for rADC ≥ 1.63 versus 6.1 months for rADC < 1.63; P = .02), whereas tumor volume, pretreatment rADC, and changes in rADC after treatment were not significantly associated with OS. Cox regression analysis revealed that post-treatment rADC significantly influenced OS (P = .02, univariate analysis), even after controlling for age and sex (P =.01, multivariate analysis), and additionally controlling for surgery after ICI treatment (P = .045, multivariate analysis). CONCLUSIONS: Elevated post-treatment rADC may be an early imaging biomarker for OS benefits in GBM patients receiving ICI treatment.

Validation of diffusion MRI as a biomarker for efficacy using randomized phase III trial of bevacizumab with or without VB-111 in recurrent glioblastoma.

BACKGROUND: Evidence from single and multicenter phase II trials have suggested diffusion MRI is a predictive imaging biomarker for survival benefit in recurrent glioblastoma (rGBM) treated with anti-VEGF therapy. The current study confirms these findings in a large, randomized phase III clinical trial. METHODS: Patients with rGBM were enrolled in a phase III randomized (1:1), controlled trial (NCT02511405) to compare the efficacy and safety of bevacizumab (BV) versus BV in combination with ofranergene obadenovec (BV+VB-111), an anti-cancer viral therapy. In 170 patients with diffusion MRI available, pretreatment enhancing tumor volume and ADC histogram analysis were used to phenotype patients as having high (>1.24 µm2/ms) or low (<1.24 µm2/ms) ADCL, the mean value of the lower peak of the ADC histogram, within the contrast enhancing tumor. RESULTS: Baseline tumor volume (P = .3460) and ADCL (P = .2143) did not differ between treatment arms. Univariate analysis showed patients with high ADCL had a significant survival advantage in all patients (P = .0006), as well as BV (P = .0159) and BV+VB-111 individually (P = .0262). Multivariable Cox regression accounting for treatment arm, age, baseline tumor volume, and ADCL identified continuous measures of tumor volume (P < .0001; HR = 1.0212) and ADCL phenotypes (P = .0012; HR = 0.5574) as independent predictors of OS. CONCLUSION: Baseline diffusion MRI and tumor volume are independent imaging biomarkers of OS in rGBM treated with BV or BV+VB-111.

Diffusion Magnetic Resonance Imaging Phenotypes Predict Overall Survival Benefit From Bevacizumab or Surgery in Recurrent Glioblastoma With Large Tumor Burden.

BACKGROUND: Diffusion magnetic resonance (MR) characteristics are a predictive imaging biomarker for survival benefit in recurrent glioblastoma treated with anti-vascular endothelial growth factor (VEGF) therapy; however, its use in large volume recurrence has not been evaluated. OBJECTIVE: To determine if diffusion MR characteristics can predict survival outcomes in patients with large volume recurrent glioblastoma treated with bevacizumab or repeat resection. METHODS: A total of 32 patients with large volume (>20 cc or > 3.4 cm diameter) recurrent glioblastoma treated with bevacizumab and 35 patients treated with repeat surgery were included. Pretreatment tumor volume and apparent diffusion coefficient (ADC) histogram analysis were used to phenotype patients as having high (>1.24 µm2/ms) or low (<1.24 µm2/ms) ADCL, the mean value of the lower peak in a double Gaussian model of the ADC histogram within the contrast enhancing tumor. RESULTS: In bevacizumab and surgical cohorts, volume was correlated with overall survival (Bevacizumab: P = .009, HR = 1.02; Surgical: P = .006, HR = 0.96). ADCL was an independent predictor of survival in the bevacizumab cohort (P = .049, HR = 0.44), but not the surgical cohort (P = .273, HR = 0.67). There was a survival advantage of surgery over bevacizumab in patients with low ADCL (P = .036, HR = 0.43) but not in patients with high ADCL (P = .284, HR = 0.69). CONCLUSION: Pretreatment diffusion MR imaging is an independent predictive biomarker for overall survival in recurrent glioblastoma with a large tumor burden. Large tumors with low ADCL have a survival benefit when treated with surgical resection, whereas large tumors with high ADCL may be best managed with bevacizumab.

Rate of change in maximum 18F-FDOPA PET uptake and non-enhancing tumor volume predict malignant transformation and overall survival in low-grade gliomas.

PURPOSE: To examine whether the rate of change in maximum 18F-FDOPA PET uptake and the rate of change in non-enhancing tumor volume could predict malignant transformation and residual overall survival (OS) in low grade glioma (LGG) patients who received serial 18F-FDOPA PET and MRI scans. METHODS: 27 LGG patients with ≥ 2 18F-FDOPA PET and MRI scans between 2003 and 2016 were included. The rate of change in FLAIR volume (uL/day) and maximum normalized 18F-FDOPA specific uptake value (nSUVmax/month), were compared between histological and molecular subtypes. General linear models (GLMs) were used to integrate clinical information with MR-PET measurements to predict malignant transformation. Cox univariate and multivariable regression analyses were performed to identify imaging and clinical risk factors related to OS. RESULTS: A GLM using patient age, treatment, the rate of change in FLAIR and 18F-FDOPA nSUVmax could predict malignant transformation with > 67% sensitivity and specificity (AUC = 0.7556, P = 0.0248). A significant association was observed between OS and continuous rates of change in PET uptake (HR = 1.0212, P = 0.0034). Cox multivariable analysis confirmed that continuous measures of the rate of change in PET uptake was an independent predictor of OS (HR = 1.0242, P = 0.0033); however, stratification of patients based on increasing or decreasing rate of change in FLAIR (HR = 2.220, P = 0.025), PET uptake (HR = 2.148, P = 0.0311), or both FLAIR and PET (HR = 2.354, P = 0.0135) predicted OS. CONCLUSIONS: The change in maximum normalized 18F-FDOPA PET uptake, with or without clinical information and rate of change in tumor volume, may be useful for predicting the risk of malignant transformation and estimating residual survival in patients with LGG.

pH-weighted amine chemical exchange saturation transfer echoplanar imaging (CEST-EPI) as a potential early biomarker for bevacizumab failure in recurrent glioblastoma.

PURPOSE: The objective of the current study was to explore the efficacy of using pH-weighted amine CEST-EPI as a potential non-invasive imaging biomarker for treatment response and/or failure in recurrent GBM patients treated with bevacizumab. METHOD: A total of 11 patients with recurrent GBM treated with bevacizumab were included in this prospective study. CEST-EPI, perfusion MRI, and standardized anatomic MRI were obtained in patients before and after bevacizumab administration. CEST-EPI measures of magnetization transfer ratio asymmetry (MTRasym) at 3 ppm were used for pH-weighted imaging contrast. Multiple measures were examined for their association with progression-free survival (PFS). RESULT: Tumor acidity, measured with MTRasym at 3 ppm, was significantly reduced in both contrast enhancing and non-enhancing tumor after bevacizumab (p = 0.0002 and p < 0.00001, respectively). The reduction in tumor acidity in both contrast enhancing and non-enhancing tumor was linearly correlated with PFS (p = 0.044 and p = 0.00026, respectively). In 9 of the 11 patients, areas of residual acidity were localized to areas of tumor recurrence, typically around 2 months prior to radiographic progression. Univariate (p = 0.006) and multivariate Cox regression controlling for age (p = 0.009) both indicated that change in tumor acidity (ΔMTRasym at 3 ppm) was a significant predictor of PFS. CONCLUSIONS: This pilot study suggests pH-weighted amine CEST MRI may have value as a non-invasive, early imaging biomarker for bevacizumab treatment response and failure. Early decreases MTRasym at 3.0 ppm in recurrent GBM after bevacizumab may be associated with better PFS. Residual or emerging regions of acidity may colocalize to the site of tumor recurrence.